w
ww
.HIMSSConference.org
#smartH
IT
FEBRUA
R
Y
1
1, 2019
ORLAND
O, FL
How
AI Enabled a Community
Hospital
to T
ackle Clinical
V
ariation
And Reduce Length-
of
-Stay
•
T
opics I w
ill
cover
•
Why use
AI for Managi
ng Clinical
V
ariation
•
T
opics I w
ill
cover
•
What we at F
lagler Hospital have done
•
T
opics I w
ill
cover
•
How we operationalized
it
•
T
opics I w
ill
cover
•
Our Results
Why reducing Clinical V
ariation Is Important
•
$750+
B wasted on
unnece
ssary ca
re
Why reducing Clinical V
ariation Is Important
•
Medical
mistakes are now
the 3
rd
leading
cause of
death in
the USA
Why reducing Clinical V
ariation Is Important
•
Now the 3
rd
leading
cause of death
in Canada
Why reducing Clinical V
ariation Is Important
•
WHO reports
95,000 d
eaths in
Europe
due t
o
Medical
Errors (1
5
th
?)
2012 Institute of Medici
ne
67%
30%
3%
Percent
Necessary
Unnecessary
Deaths
$750 B
V
ariation in Heal
thcare Expend
itures
201
1
AHA
Unex
plain
ed
T
ypes
of V
ariation
•
Common Cause V
ariation
•
Special Cause V
ariation
T
ypes
of V
ariation
•
Common Cause V
ariation
•
It
is always present
•
Special Cause V
ariation
T
ypes
of V
ariation
•
Common Cause V
ariation
•
It
is inherent
in the process
•
Special Cause V
ariation
T
ypes
of V
ariation
•
Common Cause V
ariation
•
T
o change the
results,
change
the process
•
Special Cause V
ariation
T
ypes
of V
ariation
•
Common Cause V
ariation
•
Special Cause V
ariation
•
It
does not
always happen
in the
process
T
ypes
of V
ariation
•
Common Cause V
ariation
•
Special Cause V
ariation
•
Since it is
so dif
ferent,
you would want
to ask “why”
T
ypes
of V
ariation
•
Common Cause V
ariation
•
Special Cause V
ariation
•
Eliminate or
minimize the
impact if
negative
T
ypes
of V
ariation
•
Common Cause V
ariation
•
Special Cause V
ariation
•
Not knowi
ng the dif
ference will
create wasted time & ef
fort
•
Hospitals have tried for y
ears
to reduce clinical variation
•
W
e gather
the data
we
think is
important
•
Hospitals have tried for y
ears
to reduce clinical variation
•
W
e analyze
the data then
try to implement
our
findings
•
Hospitals have tried for y
ears
to reduce clinical variation
•
By the tim
e we act,
the data
may have changed
significantly
•
Over the past several y
ears, three things have changed
•
W
e now
have massive computational
power
•
Over the past several y
ears, three things have changed
•
Our EMR has
massive amounts
of data
•
Over the past several y
ears, three things have changed
•
AI systems
like
A
yasdi can now look at
all the data and
ans
wer
questions
we did not know
to ask.
Methods of
AI
Methods of
AI
•
Supervised Learning
Methods of
AI
•
Unsupervis
ed
Learning
Methods of
AI
•
A
yasdi uses unsupervised
learning and
a branch of
mathematics called
T
opology
Methods of
AI
•
Euler
, in
1736 solution to the 7 bridge
s of Konigsberg
T
opolog
ical Map of our Pneumoni
a Pilot
Or
T
eam
•
What did w
e
need to do?
•
2,500
lines of SQL
code
to ext
ract
the data
•
What did w
e
need to do?
•
Upload to
A
yasdi
•
What did w
e
need to do?
•
Perform
Semantic and Syntactic
V
alidation
•
What did w
e
need to do?
•
Generate
the
T
reatment
Groups
•
What did w
e
need to do?
•
Select the
“Go
ldilocks” Cohort
•
What did w
e
need to do?
•
A
yasdi gener
ates
the CarePath
GL
Cohort
•
What did w
e
need to do?
•
Begin monitoring
our
providers
61.1%
8.9%
14.8%
22.3%
9.8%
28.9%
20.0%
13.3%
38.8%
31.5%
24.1%
29.1%
36.6%
23.7%
25.7%
33.3%
31.0%
12.1%
18.5%
16.5%
24.4%
34.2%
25.7%
33.3%
2.3%
6.5%
4.6%
7.8%
7.3%
0.0%
0.0%
3.3%
0.0%
1.6%
1.9%
1.0%
0.0%
0.0%
2.9%
0.0%
0%
10%
20%
30%
40%
50%
60%
70%
Cohort 216
Cohort 124
Cohort 108
Cohort 103
Cohort 41
Cohort 38
Cohort 35
Cohort 30
Comorbid Conditions
Pneumonia
Diabetes
COPD
CHF
Hypotension
Mortality
3.99
3.35
3.04
3.83
2.90
3.1
1
2.1
1
3.37
$3.17
$2.46
$2.32
$2.95
$2.42
$2.68
$2.01
$2.76
0.9%
0.0%
0.9%
0.0%
0.0%
0.0%
2.9%
0.0%
0.0%
1.6%
1.9%
1.0%
0.0%
0.0%
2.9%
0.0%
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
Cohort 216
Cohort 124
Cohort 108
Cohort 103
Cohort 41
Cohort 38
Cohort 35
Cohort 30
Cohort Ev
aluation
Pneumonia
LOS
DVC/k
Readmit
Mortali
ty
12.30
3.49
4.91
2.89
$13.05
$4.21
$5.10
$3.21
1.1%
0.0%
1.6%
0.0%
9.0%
34.8%
9.4%
24.6%
0%
5%
10%
15%
20%
25%
30%
35%
40%
0.00
2.00
4.00
6.00
8.00
10.00
12.00
14.00
Cohort 8
9
Cohort 6
9
Cohort 6
4
Cohort 5
7
Cohort
A
nal
ysis
Septic Shock
LOS
DVC/k
Readmit
Mortali
ty
$35,310.79
$17,634.29
$49,316.00
$171,135.6
4
$(74,951.68
)
$(100,000
.00)
$(50,000.
00)
$-
$50,000.
00
$100,000
.00
$150,
000.00
$200,000
.00
Savings B
ooked to
Date
T
o
t
a
l
S
a
v
i
n
g
s
$198,445.04
Pneumonia
COPD
CHF
Septic Shock
Sepsis w/o Shock
•
What do y
ou need to do
•
Need the SQL
skills to retrieve the data
•
What do y
ou need to do
•
Bring phy
sicians
in early
•
What do y
ou need to do
•
Recogni
zed that it is an iterative process
•
What do y
ou need to do
•
W
ork hard
•
What do y
ou need to do
•
W
e can change
the worl
d!
w
ww
.HealthcareMachineLearningAI.com
#smartH
IT
Michael C. Sanders, M.D.
Flagler
Hospital
Michael.Sanders@FlaglerHospital.org